Re: I found a FORTH stack optimzer, but could use a better version.

"Marcel Hendrix" <mhx@[EMAIL PROTECTED]
> wrote in message
news:38331417223559@[EMAIL PROTECTED]
> "Rod Pemberton" <do_not_have@[EMAIL PROTECTED]
> writes Re: I found a FORTH
stack optimzer, but could use a better version.
>
> > I ran across an interesting optimizer for FORTH stack words in java by
Ram
> > Janovski, and Vassilii Khachaturov, dated 2004, here:
>
> >  http://black.cs.bgu.ac.il/forth-opt
>
> > Depending on which FORTH stack words you select, it'll attempt to
generate a
> > optimized definition using the input set of words.
>
> I fail to see what's being "optimized" here. In what sense is the result
> "optimal?" It seems to help in writing code that never should get
written
> in the first place, or that should use locals (e.g. when writing for a
GUI
> where this style can't be prevented).
>

I think their goal, on one of the linked to pages, was to find the minimal
sequence of FORTH words for a given stack conversion using some algorithm
of
Dijkstra's.  E.g., when all sup****ted words are enabled (I wasn't doing
this...), to go from (ab) to (ababba), or (ab -- ababba), their result is:

  Transforming ab ==> ababba
  2DUP ( ab -- abab )
  2DUP ( abab -- ababab )
  SWAP ( ababab -- ababba )

Sorry, I wasn't using their optimizer for FORTH, although that's
(apparently) what it was written for.  My use is for a optimizing a
compiler, but it could be useful for a FORTH implementor to optimize the
underlying assembly...  Originally, I was looking for ways to reduce
FORTH-like stack operations to simple operations which could correspond to
x86 assembly for a single stack, not a double stack environment like
FORTH.
So, I was checking how each primitive, word, or verb, could be
(re)written.
E.g., DUP, DROP, OVER, PICK, "PLACE", >R, R>, R@[EMAIL PROTECTED]
 etc. correspond nicely
to
x86 instructions, but other ones necessary for stack operations, such as
ROT, TUCK, SWAP, ROLL don't.  Those operations relocate or move a block of
data on the stack, even if small.  I'd like to avoid them.  SWAP
corresponds
nicely for register operations, but not stack operations...  So, I was
using
their  stack optimizer to see how each primitive could be written using a
small set x86 compatible primitives, words, verbs, like these:

  1) >R,R>,R@[EMAIL PROTECTED]
  2) DUP,DROP,OVER
  3) DUP,DROP,TUCK
  4) DUP,DROP,ROT
  etc.


[links]

Thanks for the links.  I'll take a look.


Rod Pemberton